MCP server for AI agents

These docs are also served as an MCP server. Connect it to the AI agent that writes your integration (Claude Code, Cursor, or any MCP client), and the agent can read the guides, the rules it must follow, and the exact request and response shapes of every operation, instead of guessing them.

Setting	Value
URL	`https://admin.ramp.tucop.xyz/mcp`
Transport	Streamable HTTP, stateless, JSON responses
Auth	None. The docs are public.
Access	Read-only. It never calls the TuCOP Ramp API and never needs or accepts an API key.

Connect

Claude Code

bash

claude mcp add --transport http tucop-ramp-docs https://admin.ramp.tucop.xyz/mcp

Cursor and other clients that take a JSON config (for Cursor, .cursor/mcp.json):

json

{
  "mcpServers": {
    "tucop-ramp-docs": {
      "url": "https://admin.ramp.tucop.xyz/mcp"
    }
  }
}

Clients that only speak stdio can use an adapter such as npx mcp-remote https://admin.ramp.tucop.xyz/mcp.

Tools

Tool	What it returns
`get_agent_rules`	What an agent must and must not do when integrating TuCOPRamp. Agents are told to call it first.
`list_docs`	Every page: slug, title and description, in navigation order.
`read_doc`	One page as markdown, by slug (for example `authentication`).
`search_docs`	Pages that match some words, with short snippets. Arguments: `query`, optional `limit`.
`list_operations`	API operations: operationId, method, path and summary. Optional `tag`.
`get_operation`	One operation with parameters, request body and responses, as YAML. By `operationId`, or by `method` and `path`.
`get_integration_guide`	The guide pages for an integration model: `own_provider_account` or `per_user_kyc`.

Every page is also a resource, tucop-docs://<slug> (markdown), and the API spec is tucop-docs://openapi.yaml.

Try it with curl

bash

curl https://admin.ramp.tucop.xyz/mcp \
  --header "Content-Type: application/json" \
  --header "Accept: application/json, text/event-stream" \
  --data '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"read_doc","arguments":{"slug":"quickstart"}}}'

The content is the same as this site, updated with each deploy. What the server answers is documentation only: your agent still has to test against the sandbox with a test key before anything goes to production.